Nala Assistant

The project is designed for quick prototyping — a basic web-enabled voice assistant can be deployed locally in minutes using the provided FastAPI backend and front-end. Full model capabilities (e.g., Bark or local LLMs) may require additional setup such as GPU resources.

Standard FastAPI deployments apply; however, running local LLMs or high-quality TTS models typically requires an NVIDIA GPU (examples referenced include V100/A100). The project supports typical server setups with SSL and process managers like gunicorn/uvicorn.

The codebase demonstrates integration with Whisper-based transcription, SpeechT5 and Bark for TTS, and Dolly-v2 for LLM responses. The architecture includes helper functions to add or swap other models.

The project is licensed under the Apache License 2.0.